Deep Learning for Transesophageal Echocardiography View Classification

This research article presents a study on the development of a deep learning-based model for classifying transesophageal echocardiography (TEE) views. The primary objective of the study was to create a multi-category TEE view classification model that adds structure to the unstructured data obtained from intraoperative and intraprocedural TEE imaging. A convolutional neural network (CNN) was trained using labeled TEE videos from Cedars-Sinai Medical Center, and the model was externally validated with videos from Stanford University Medical Center. The study reports high accuracy across various standardized TEE views, with the best performance noted for specific views such as the Trans-Gastric Left Ventricular Short Axis View and the Mid-Esophageal Long Axis View. The findings suggest that the deep learning model can effectively classify TEE views, potentially facilitating further deep learning analyses in the context of intraoperative and intraprocedural TEE imaging. The research underscores the importance of TEE imaging in managing complex cardiovascular diseases and highlights the potential of AI in enhancing echocardiography practices.